# Linear Voltage Regulator Quirk Using the TL072/4

I recently designed this simple voltage regulator circuit using a TL072 on LTspice (and multisim)

And the circuit worked as expected: it output 5V.

But when I built it on the breadboard, something weird and counterintuitive happened. The output of the regulator was actually 0V! When I checked the output of the op amp, it was around 14V. Now here's the funny thing. The inverting input was at 2.495V, as expected, but the non inverting input was at 0V. Despite this, the output of the op amp was 14V!

Next, I replaced the TL072 with an LM358. And guess what happened? It worked!!

Now what I want to know is why this circuit worked with the LM358 (but not the TL072) on the breadboard when LTspice and Multisim simulations clearly showed that the circuit works with the TL072.

PS to get the circuit working with the TL072, I actually had to temporarily short the inverting and non-inverting inputs to get the circuit going. But that is not a proper solution. Also, the LM358 is inferior in performance to the TL072, which is why I want to find a solution and make the circuit work with the TL072.

Quite simply, the TL opamps are not single supply devices, their common mode don't reach to the V- supply and neither their output. Simply they aren't made for that kind of supply, while the 358 is.

The 'proper solution' would be of course to use a regulator like the 317, but I assume this is an educational experiment.

Assuming you are powering the op. amp. with the unregulated voltage:

Input voltage:

• if you are using the TL072H, you are close to the limit, but within spec. (2V)
• if you are using any other TL072* suffix, you are out of worst case spec. (4V)

Output swing: the diodes give you enough headroom to even turn off the PNP BJT (with the TL072H, but not with the others). The lower limit (close to 0V) is not relevant with your topology (for 15V V+ op. amp. supply).

I'm also assuming this is just an educational experiment, but also wonder why you chose this topology instead of one with a NPN BJT.

Have you analyzed the response to input ripple and load change? Also, have you considered what happens if you remove the 100n feedback capacitor?